Jason's Favorite Sessions (Not Yet Reserved)

When one event impacts hundreds, or even thousands, of customers simultaneously (a cancelled flight, a power outage, a system failure), contact centers cannot scale human agents fast enough. In this hands-on workshop, you orchestrate parallel AI agent teams using Amazon Connect Customer. You configure AI agents that fan out across dozens of affected customers at once: rebooking flights, issuing credits, and scheduling callbacks, all while a single human orchestrator monitors progress, approves exceptions, and intervenes only where needed. Walk away with a working parallel orchestration pattern and the configuration to deploy 1-to-many AI response at scale. This is not 1-to-1 AI coaching. This is 1-to-many crisis orchestration.

Today's AI agents forget everything after each conversation ends. A customer calls back about the same issue for the third time, and the agent starts from zero. In this chalk talk, we explore how to architect persistent memory systems for Amazon Connect Customer AI agents, enabling them to remember customer preferences, recall prior issue context, and adapt resolution strategies over time. We walk through three memory patterns: episodic (past interactions), semantic (customer knowledge graphs), and procedural (learned resolution paths). We whiteboard real architectures using Customer Profiles, Bedrock knowledge bases, and DynamoDB to build agents that reduce repeat contacts and deliver personalized experiences, without retraining models. Bring your memory challenge and we'll design the persistence layer on the whiteboard.

Every analytics query starts with how your data is stored. In this session, learn how Amazon S3 Tables deliver purpose-built Iceberg table storage that accelerates query performance, reduces storage costs, and eliminates table maintenance overhead. See how to pair S3 Tables with Athena, Redshift, and EMR to get faster results at lower cost, automate compaction and snapshot management, and simplify your lakehouse without changing your query engine or workflow.

Join us for a collaborative learning exercise to develop practical skills in platform engineering and cloud-native architecture on AWS. Dive into a live AWS environment where you will build and operate internal developer platforms, design cloud-native applications, and implement container orchestration at scale. Build skills in containerization, container orchestration platforms, service mesh, platform APIs, and self-service infrastructure as you construct golden paths that empower development teams. Face scenarios that challenge your ability to balance developer experience with operational governance! Whether you are a seasoned platform engineer or just exploring cloud-native patterns, this GameDay will put your skills to the ultimate test. Come solo or bring a group. Either way, you'll compete on a team. This session features integrated use cases with Red Hat. Walk-ups welcome.

An AI agent completes five steps, waits for approval, and then its runtime is replaced. How does it recover without repeating costly tool calls or losing the business process? In this session, learn how Amazon Bedrock AgentCore, Strands Agents, and Temporal combine AWS-native agent infrastructure with durable execution. We’ll build a production agent that preserves progress across failures, coordinates tools, agents, and people, and safely evolves while work is still running. You'll see how Temporal natively supports serverless compute infrastructure, complementing AgentCore Runtime with durable execution, preserving workflow progress across failures, approvals, runtime changes, and scale to zero, without pushing recovery and coordination logic into application code. This presentation is brought to you by Temporal, an AWS Partner.

AI agents generate SQL, surface insights, and take action on your data. Without governance guardrails, they hallucinate column meanings, query stale tables, and return wrong answers with high confidence. In this session, learn how to ground your AI agents in trusted metadata using AWS Glue Data Catalog. See how to flow business context like glossary terms, descriptions, and quality scores into agent workflows so they generate accurate, schema-aware queries instead of guessing. Walk away with practical patterns for governing AI-assisted analytics the same way you govern human analysts

Come with your team of 2 or join one on the spot. Compete head-to-head, race the leaderboard, and win prizes. Solve real-world Gen AI and Agentic AI scenarios using Amazon Bedrock, SageMaker AI & Kiro - including a sponsored challenge from NVIDIA to build and optimize GenAI workloads using NVIDIA's accelerated computing and AI inference stack on AWS. Challenges suit all skill levels, with AWS experts and helpful clues throughout. All you need: a laptop, curiosity, and a competitive spirit.

Open table formats and open-source engines like DuckDB give organizations the freedom to choose the right tool for each workload, from interactive dashboards and exploratory analytics to large-scale data processing, with no lock-in. In this session, learn how to capitalize these open tools to build on AWS and turn it into a competitive advantage for your business. We deep dive into the application of each engine, how they complement one another across the analytics lifecycle, and how data governance capabilities like catalog access, fine-grained governance, and consistent table management on AWS further simplify multi-engine architectures in production.

If you're running Express, Hono, or Fastify on Lambda today, you're wrapping your HTTP server in an event-handler adapter. Lambda Web Functions drops that requirement. Your code listens on a port, same as localhost. Multi-concurrency keeps sandboxes warm across requests, delivering consistent response times. Active-duration billing charges for CPU consumed, not time spent waiting on a database response. Opt-in multi-region replication puts your app closer to users: pick your regions, code distribution handles the rest. This session covers the execution model, billing mechanics at real workloads, multi-region setup, and a live deployment from zero to global endpoint.

The demo shipped. The agent went live. Then someone pulled it back. 74% of organizations that have deployed an AI agent have rolled it back or shut it down, and the rate rises to 81% among those reporting fully mature guardrails. Getting agents into production got easier. Keeping them there without hurting customers did not. We'll walk through the failure patterns behind rollbacks: agents that fail silently, prompts treated as safety boundaries, and monitoring that overlooks non-deterministic failures. You'll leave with the patterns that prevent them: infrastructure-level guardrails, blast radius containment, and observability built for probabilistic systems.